Other packages that would save me from doing this work?

[sfirke]

questionr has no vignette so hard to tell. I like that it says it uses Shiny, but where can I see something about that? And functions like cprop don't work with data.frames, if that's true for the underlying data classes and other functions this may not be tidyverse-friendly.

memisc?

Could try surveydata but the limitations noted on its README seem undesirable. I think better to not invent a new class of object, and instead work around limitations of data.frames if possible.

[sfirke]

Ooh maybe the labelled package has the right idea: use the labelled class from haven. Here's the vignette.

[sfirke]

I wrote an S3 method for tabyl on a labelled vector in a normal data.frame.

tabyl.labelled <- function (vec, sort = FALSE, show_na = TRUE, ...) {
  var_text <- labelled::var_label(vec)
  vec <- to_factor(vec)
  res <- janitor::tabyl(vec, sort, show_na, ...)
  names(res)[1] <- var_text
  res
}

That drops the var_label into the 1st name of the tabyl output:

> tabyl(sav_responses$q0001)
               What is your current role on your campus?   n     percent
1              Other, not administrator (please specify)  13 0.020280811
2                                              Principal 145 0.226209048
3                                    Assistant Principal 223 0.347893916
4                                                   Dean  41 0.063962559
5                                                  Coach 214 0.333853354
6                                   Administrator, other   0 0.000000000
7                        Classroom teacher (not a coach)   5 0.007800312
8        I do not currently work in XXXXX Public Schools   0 0.000000000

The annoyance is that actually, a column can have a var_label without being class labelled. So an S3 method doesn't work. Would there be harm in coercing all variables in a data.frame to be labelled, for this purpose? Could there be a leaner class that just has var_label attribute, and we add that as a class to every var_label'ed column in the data.frame?

[sfirke]

Maybe some mix of Hmisc and labelled packages. Hmisc::label() does set class = labelled when called which I need for S3. But I like the interface of the labelled package better.

So maybe the data storage paradigm is, labels in the label attribute, then create tabyl, crosstab, and check-all-that-apply S3 methods. Value labels might be trickier, but they could pull the var label and set it as the column name before calling the regular tabyl.data.frame.

W/ the current SurveyMonkey process, it would be nice to set all labels at once. I can't find a function that lets me supply a character vector to a data.frame, unbelievably. I guess I'll wrap my StackOverflow answer into a function?

[sfirke]

I could use this approach to create a named list, then pass that to labelled::var_labels - but I don't see how that's any better than my for loop linked on StackOverflow above.

[sfirke]

There's also now the labelVector package: https://cran.r-project.org/web/packages/labelVector/vignettes/labelVector.html. Designed for use in other packages, and labelVector::set_label does set class to "labelled"!

[sfirke]

There is also sjlabelled on CRAN. Gah.

[sfirke]

Update: now, using janitor v1+ and setting labels with labels(dat) <- char_vec using papeR::labels, the functional prototype is:

tabyl.lv <- function (vec, show_na = TRUE, ...) {
  var_text <- attr(vec, "variable.label")
  vec <- labelled::to_factor(vec)
  res <- janitor::tabyl(vec, show_na, ...)
  names(res)[1] <- var_text
  res
}